Urban air quality has become a critical concern due to rising pollution levels and their direct impact on public health and environmental sustainability. This study presents a comparative analysis of various machine learning models aimed at forecasting fine particulate matter (PM2.5) concentrations and overall Air Quality Index (AQI) values in the context of smart city infrastructure. The models evaluated include Linear Regression, Support Vector Regression (SVR), Random Forest, XGBoost, and Long Short-Term Memory (LSTM) networks. Historical air quality datasets sourced from public environmental monitoring agencies were used, covering a diverse range of meteorological and pollutant features. Evaluation was conducted using standard performance metrics such as Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and the Coefficient of Determination (R² score). Among the tested models, XGBoost consistently demonstrated superior accuracy in both PM2.5 and AQI predictions, attributable to its robustness against outliers and efficient handling of non-linear data patterns. The results underline the practical applicability of advanced ML models in building predictive air monitoring systems that can be integrated into smart city platforms for proactive environmental management and policy-making.

Keywords: PM2.5, AQI, Machine Learning, Air Quality Forecasting, Smart Cities, Urban Pollution, Environmental Data